May 2, 1997	-1-

Fixes for JMP Versions 3.1.6 to Version 3.2.1

---------------------------------------------------------------------------------------------------
Fixes Previously Completed for JMP Versions 3.1.6.1, 3.1.6.2,  and 3.1.7:

  Correlations of constant columns correctly show as 0.

  In a logistic regression, the number of iterations has been increased before the process   stops.
  
  The Macintosh 68000 without Color Quickdraw no longer causes problems.

  Windows95 JMP windows maximize correctly with no problems.  

  There is a more flexible policy on JMP.PER for write-protected directories. 

  The Import dialog recognizes labels when the Labels radio button is checked.

  Rotated text on '030 Macs no longer causes problems.

  Column Names which wrap while typing into the column header leaves display correctly.

  The New Column and Column modification dialogs correctly show the current format type a checked item.

  The New Column and Column modification dialogs show the current date and date increment from the Format popup menu as checked items.

  The axis modification dialog now automatically formats the minimum and maximum with Best format as determined by the data.

--------------------------------------------------------------------------------
Additional Fixes for JMP Version for 3.2:

  The JMP installer will not install JMP until you have installed Win32S, if you need Win32S.

  All known problems with maximizing JMP report windows under Microsoft Windows have been fixed.

  Windows now live-drag and resize.

  The JMP installer has been modified so that it no longer attempts to install Win32s.  Rather, we alert the user if Win32s is not installed or is an earlier version and tell them to install Win32s outside of the JMP install before continuing with the JMP installation.  The JMP install will exit after informing the user.

  On the File Import Dialog (Windows only) you can edit the  "Other" checkbox to include any characters, as well as  hex  values (like 0x45).

  The File Import dialog has a separate Options dialog so that the dialog does not get cut off on small screens.
  
  The File Import Text dialog preserves the settings you had previously set.

  Importing data no longer creates a missing column at the end of the data table.

  A problem saving a data table with an instrument data source column on a powerbook has been fixed. In the event that there is a problem, a Saved Failed error message is displayed. 

  In the Data Table, if you type in character into a new numeric column, JMP offers to change the column from numeric to character, rather than just complaining.

  In the Data Table, the number of selected rows and columns  now shows in the lower left hand corner of the spreadsheet (left of the bottom scroll bar).

  In the Data Table, 2 clicks in the different areas of the spread sheet sometimes are no longer mistaken to be a double-click on the add default col area.

  In the Data Table,  List Check now works across concatenation; it is the union of values.

  In the Data Table, a fuzz value has been added to Range Check bounds.

  In the Data Table, making global column changes no longer causes a crash. 
  
  In the Data Table, there is a Refresh Data Type button in the Col Info dialog when the Data Source selection is  Instrument . 

  A By-group table name can be as long as a standard JMP data table name.

  A By-group table with By-Mode on no longer crashes with its window is maximized or minimized.

  The Calculator no longer complains about a recursion error when you reorder columns.

  The Calculator constant entry field is wider to enter or edit long character strings

  The Calculator constant entry field will now autoscroll text correctly.

  The Calculator has internal improvements, especially to make derivatives cleaner and faster,  with less duplicate recalculating.

  The Calculator date formula date(day 0, month m, year y) returns the same as date(day 1, month m, year y) - 1 day, instead of "undefined results".

  In the calculator, when columns have the same name for a label but differ in capitalization, the first appearance of the column name replaces the other similar column names in formulas after the data table has been saved and reopened.
  
  The Calculator now correctly handles the Std, Sum, and Product functions when all values are missing.
  
  The Calculator correctly uses the noncentrality parameter to evaluate in the Chi-Square Density Function.
  
  The Calculator beta and quantile functions evaluate correctly (they were returning the density instead of quantile).
    
  The Calculator applies stricter parameter checking to all probablility functions.
   
  The Calculator has many new quantile, density, and distribution functions (see the Whats New document).  

  Under Windows, the Calculator handles empty parameters for character functions; namely, munger, upper, lower, length and others.

  Under Windows, the Calculator new parameter dialog no longer crashes when you change the data type from numeric to character and click OK. 

  Under Windows, scrolling horizontally when an analysis is in a maximized state no longer causes the next analysis window to be opened in a similarly scrolled state. 

  Under Windows NT using the char2date function on a null string no longer crashes.

  Group/Summary command computes weighted standard deviation and standard error using ( n -1) instead of 
(sum weights -1). 

  The Group/Summary command includes computations of a range statistic.

  In Distribution of Y histogram bars are automatically narrowed if they do not project at least halfway to the end of the range. This improves the appearance of the graph for spiky distributions as when there are only a few integer values.

  In Fit Y by X for continuous X and continuous Y, selecting Plot Residuals several times removes old plots before showing new plots.  

  Fit Y by X with Continuous Y and  Nominal X correctly adjusts the degrees of freedom when all responses for a level are missing.

  The Fit Model Standard Least Squares platform offers the Press statistic option on the Whole Model report. This is the sum of squares residuals where each residual is calculated after dropping that row from the estimate. 
  
  Fit Model Standard Least Squares handles large scale terms correctly instead of reporting nontestable contrasts when the scaling of the variables was large.

  Fit Model Standard Least Squares singularity checking in a custom test now adjusts to the scale. Previously, large regressor values caused a spurious declaraction of singularities in custom test.

  Fit Model Screening platform has a popup option that allows you to turn off the mesh surface plots in the contour profiler.

  The profiler in the Fit Model Screening platform works with desirability function even when the model is not fully estimable. 
 
  Clicking the last term in a mixture model in the contour plot for response surface report (not the contour profiler)no longer causes problems.

  The  Student's t distribution (quantile) is now accurate for more extreme values.

  Certain dialog edit boxes have been made wider so that the numbers appear whole, rather than clipped.
 
  Clicking to the left of the first point in an EWMA chart no longer causes problems.

  A Pareto Chart bar when clicked correctly reflects values of classification (x) variables.
  
  A Pareto Chart bar label is correct when bars are combined.

  In Shewhart charts, the box plots no longer hit the edge of the plot.

  You now get the appropriate IO error message while reading/writing JMP tables (e.g. disk full problem).

  On the Survival platform, the Exponential, Weibull, and Lognormal plots are calculate  using the survivals adjusted to use the non-zero times, which produces the correct plots.

  On the Survival platform, the y-axis range has been expanded to better accommodate the results.
  
  On the Survival platform, group names can be longer.

  During Nonlinear computations, a check for a cancel signal is made every 10 observations. 

  During Nonlinear computations, when in irregular areas of the likelihood where the curvature wasn't nonpositive definite, a text window logging this was put up. This obscured the output and was of little use, so it has been removed.

  Previously in Nonlinear fitting, if you excluded points after one fit, it couldn't get started on the next fit because it confused exclusion with having more missing points. The iteration check for improving the fit always requires that you do not lose observations, such as going out of range of functions. This has been fixed.

  In Fit Nonlinear the intermittent 2nd deriv crash is fixed.

  In Fit Nonlinear the calculator runs up to seven times faster for some formulas.

  Logistic regression improved responsiveness to Cancel  buttons that allow you to escape quicker when inadvertently starting a very large problem. Now it checks at every row.

  When journaling, the occasional missing minus signs in formulas now correctly appear. 

  In an Applescript "Saving Yes" works on the Close command.

--------------------------------------------------------------------------------
Additional Fixes for JMP Version for 3.2.1:

  When you shut down your computer, you are prompted for saves. 

  The Import statement can now read unpacked FACS data.

  When a column is changed to Integer data type, the field width automatically adjusts to accommodate the maximum value. 

  The Import statement strips single quotes when importing text.

  All delimiters (semi-colon and others) are working for Import under Windows.

  For Short Numerics, entering the out-or-range value (-127, -32767, or -2147483647) now  generates an error message.	

  Range Check for columns is now working properly. 

  When you Concatenate tables that have List Check columns, the new List Check column contains all values found in both tables, instead of just those found in the resulting column itself.

  Trailing blanks have been removed from character variables when using the Open command to read SAS data sets.

  Under Windows, saving a journal with curly braces (}) now creates a valid RTF file. Curly braces sometimes occur in Fit Model Stepwise with a nominal effect. 

  Under Windows, the status bar text stringfor the Tables menu items is now correct.

  Problems with deleting a column subscript in the Calculator Sum and Product functions are fixed.

  The Magnify tool now works correctly in the upper range of a Distribution of Y histogram.

  The matching analysis in Fit Y by X, now correctly calculates the p-value for the interaction.

  In Fit Model Standard Least Squares the option to choose the Press Statistic no longer appears where is unavailable.  

  The Fit Model Screening analysis now computes the correct slope for line of fit when using Lenth's PSE in the normal plot.

  The Fit Model Screening analysis is no longer sensitive to whether or not the Profiler preference is set.  

  In Shewhart Charts, if probability is specified instead of k sigma, the correct lower limit for R or S is computed.